lichess.org
Donate

Grandmaster game - endgame question

I bought the book Winning Chess Endings by GM Seirawan.
It features a game Cruz-Seirawan from the Olympiad of 1994.

The endgame is quite exordinary. Seirawan writes that his team mates were shocked, and thought that his 48...Ra5+ was a blunder and would lead to a draw.
It is winning however!

In the book Seirawan gives several exclamation marks to his moves after that, but Stockfish claims 54...e4 was a serious mistake, dropping the evaluation from -7.9 to -1.2 although "without comment" (hmmm ???) a few moves later it increases suddenly to -9.3 (!!).

What do you think about the Stockfish analysis ?
I seem to have read that Stockfish on Lichess takes 4 seconds per move to analyse. That might not be enough for difficult pawn endings ?

Does someone have a more extensive (chess engine) analysis available ?

Fascinating endgame by the way, a "must see" game.

Also here, with some user comments below :
http://www.chessgames.com/perl/chessgame?gid=1459045
Thumbs up to Yasser for calculating this (assuming this is actually won for Black).

Not sure about this ending, but i can tell you for sure that Stockfish on Lichess is useless for analyzing deep endgame lines. The engine just seems to be cut off at some point (after fix 4 seconds maybe?), even when calculating a crucial line where it would need to go a bit deeper. At least for elder people like me this brings back good, old "horizon" effect from the early years of computer chess.
Erm, considering extensive analysis: Just start your database, open the game, start Stockfish, wait a minute, look at the evaluation?
The move e4 doesn't spoil the win, but the computer prefers a different pawn breakthrough. after 53. Kd3 f5 54. Kc3 another way to win is: 54...h5 55. Kb3 f4 56. Kc3 f3! 57. Kb3 e4 58. Kc3 e3 (or 57. Kd3 Kb4 (outflanking) 58. Ke4 Kxc3 59. Kxe5 Kd3 when the f-pawn queens first)
#3 I agree; without Syzygybases (tablebases) installed Stockfish fails to analyze complex endgames... computers in general take a long time to analyze complex endgames even with tablebases.
e4 is the top 6 choice of SF. But all these moves are winning.

1 43 -123.43 54.... f4 55.Kd3 f3 56.Kc3 e4 57.Kd2 Kxc4 58.Ke3 Kc3 59.Kxe4 Kd2 60.Kf4 h5 61.Kg5 Ke2 62.Kf5 Kxf2 63.Kf4 Kg2 64.Kg5 Kxh2 65.Kf4 f2 66.Ke3 Kxg3 (74.86)

2 43 -123.42 54.... h5 55.Kd3 f4 56.Kc3 f3 57.Kd3 e4+ 58.Ke3 Kxc4 59.Kxe4 Kc3 60.Ke3 Kc2 61.Ke4 Kd2 62.Kf4 Ke2 63.Kg5 Kxf2 64.Kf4 Kg2 65.Kg5 Kxh2 66.Kf4 f2 67.Ke3 Kxg3 (74.86)

3 43 -123.41 54.... h6 55.Kb3 h5 56.Kc3 f4 57.gxf4 exf4 58.Kd3 f3 59.Kc3 h4 60.Kd3 Kb4 61.c5 Kxc5 62.Ke3 Kc4 63.Ke4 Kc3 64.Ke3 Kc2 65.Ke4 Kd2 66.Kf4 Ke2 67.Ke4 Kxf2 68.Kd3 Kg2 69.Kc2 Kxh2 (74.86)

4 43 -123.40 54.... Kb6 55.Kc2 h5 56.Kd3 Kc6 57.Ke3 Kc5 58.Kd3 f4 59.Kc3 f3 60.Kd3 e4+ 61.Ke3 Kxc4 62.Kxe4 Kc3 63.Ke3 Kc2 64.Ke4 Kd2 65.Kf4 Ke2 66.Kg5 Kxf2 67.Kf4 Kg2 68.Kg5 Kxh2 69.Kf4 f2 70.Ke3 Kxg3 (74.86)

5 43 -123.40 54.... Kc6 55.Kc2 h5 56.Kd3 Kb6 57.Ke3 Kc5 58.Kd3 f4 59.Kc3 f3 60.Kd3 e4+ 61.Ke3 Kxc4 62.Kxe4 Kc3 63.Ke3 Kc2 64.Ke4 Kd2 65.Kf4 Ke2 66.Kg5 Kxf2 67.Kh4 Kg2 68.Kg5 Kxh2 69.Kf4 f2 70.Ke3 Kxg3 (74.86)

6 43 -123.40 54.... Kd6 55.Kb4 Kc6 56.Kc3 Kc5 57.Kd3 h6 58.Kc3 f4 59.Kd3 f3 60.Kc3 e4 61.Kd2 Kxc4 62.Ke3 Kc3 63.Kf4 Kd3 64.Kf5 e3 65.fxe3 Kxe3 66.Kg6 f2 67.Kh5 f1=D 68.Kg6 Dg2 69.h4 gxh3 70.Kg7 Dxg3+ (74.86)

7 43 -123.40 54.... e4 55.Kb3 h6 56.Kc3 h5 57.Kb3 f4 58.gxf4 e3 59.fxe3 h4 60.f5 Kd6 61.Kb4 Ke5 62.c5 Kxf5 63.Kb5 g3 64.c6 Ke6 65.h3 g2 66.Ka6 g1=D 67.c7 Kd7 68.Kb6 Dxe3+ 69.Kb7 Db3+ 70.Ka6 Dxh3 (74.86)
I am not 100% sure, but I think lichess only uses a "mini version" of Stockfish, or an old one. Maybe there are better versions of Stockfish than the one used by Lichess. I might be wrong.
Stockfish analysis on this site is only 2 seconds per move maximum. For a deeper analysis, install a chess program and analyze the game there.

This topic has been archived and can no longer be replied to.